Agent capability
Agent Authorization, with verifiable intent.
Grant agents narrow, time-boxed permission to act, never blanket access. Every action carries a signed, scoped mandate you can verify, audit, and revoke in real time.
{"agent_id": "agt_9f2c...","scope": ["cart.read", "checkout.write"],"constraints": {"max_amount": "250.00","currency": "USD","expires_in": "300s"},"intent": "purchase_approved_by_user","result": "mandate_issued"}
What is agent authorization?
Agent authorization is the practice of granting an AI agent narrow, time-boxed, revocable permission to act, never blanket or standing access. With LoginID, every action carries a signed, scoped mandate bound to verified human intent: you define exactly what the agent may do (for example, read a cart and complete a checkout up to a set amount), for how long, and you can verify, audit, or revoke it in real time.
Least-privilege access for autonomous actors.
Fine-grained scopes
Authorize exactly what an agent may do: read a cart, complete a checkout, nothing more. No standing access, no over-broad tokens.
Verifiable intent
Each action is bound to a signed mandate that captures user intent, so you can prove what was approved and why it happened.
Time-boxed & revocable
Mandates expire on a schedule you set and can be revoked instantly, limiting blast radius if an agent goes rogue.
Full audit trail
Every authorization decision is logged and attributable, giving risk and compliance teams a clean record of agent activity.
How agent authorization works.
- 01
Request a scoped mandate
An authenticated agent requests permission for a specific action with explicit limits: amount, surface, and lifetime.
- 02
Bind to verified intent
The mandate is tied to a passkey-confirmed human approval, capturing exactly what the person authorized.
- 03
Enforce and audit
Your app honors the scoped mandate, expires it on schedule, and keeps a verifiable record of every decision.
Human Authorization
Let real people approve exactly what their agents may do with a single tap: clear scopes, clear limits, full control.
Learn more- Approve with a passkey
- See and set scopes
- Revoke access anytime
Frequently asked questions
- What is a scoped mandate for an AI agent?
- A scoped mandate is a signed authorization that grants an agent permission for a specific action with explicit limits, such as a maximum amount, a single merchant, and a short expiry. It captures the verified human intent behind the action so you can prove what was approved and enforce least privilege.
- How do you limit what an AI agent is allowed to do?
- LoginID issues fine-grained scopes (for example cart.read or checkout.write) instead of broad tokens. Each scope is time-boxed and revocable, so an agent can only perform the exact actions a verified human approved, within the limits you set.
- Can an agent's permissions be revoked?
- Yes. Mandates expire automatically on a schedule you define and can be revoked instantly, which limits the blast radius if an agent behaves unexpectedly or is compromised.
- How is agent authorization audited?
- Every authorization decision is logged and attributable to a verified human and a specific mandate, giving risk and compliance teams a clean, verifiable record of agent activity.
- How is this different from OAuth scopes?
- OAuth scopes are typically long-lived grants tied to an app. LoginID mandates are short-lived, single-intent, cryptographically signed, and bound to a verified human approval, purpose-built for autonomous agents acting on a person's behalf.
Grant intent, not blanket access.
Issue narrow, signed, revocable mandates so agents act only within the limits a verified human approved. Ship least-privilege agent authorization in days.